GPT-5.2 Codex
GPT-5.2 Codex is a coding-oriented chat model from OpenAI, part of the Codex line of models tuned for software engineering and agentic development work.
API Pricing
| Provider | Input / 1M | Output / 1M | Cached / 1M |
|---|---|---|---|
| $1.75 | $14.00 | $0.175 |
Prices updated daily. Last check: Oct 14, 2026
Compare API pricing for every OpenAI model →GPT-5.2 Codex pricing by provider
Input, output, and batch rates, plus alternatives, for one provider at a time.
Performance & Benchmarks
Source: Artificial Analysis →Reasoning & Knowledge
- GPQA Diamond89.9%
- Humanity's Last Exam35.7%
Agentic & Tool Use
- Terminal-Bench Hard37.1%
- τ²-bench92.1%
Instruction & Long Context
- IFBench77.6%
- Long-Context Reasoning82.3%
Benchmarks measured Oct 2026. Scores are independent evaluations, not vendor-reported.
Model Details
General
- Creator
- OpenAI
- Modalities
- Text
Capabilities
- Open Source
- No
Strengths & Limitations
Strengths
- Part of OpenAI's Codex line, which is positioned specifically for software engineering and code-editing workloads rather than general chat
- Exposed through OpenAI's standard chat-completions style interface, so existing OpenAI client code and SDKs apply
- Sits in the GPT-5 series numbering, making it straightforward to A/B against sibling GPT-5.x models in the same API
- Suited to agentic coding loops where the model repeatedly reads context and proposes edits
- Available from multiple listed providers where applicable, so per-token rates can be compared side by side in the table on this page
- Chat-format serving means it can be dropped into existing prompt-and-tool orchestration frameworks without a separate integration path
Limitations
- We do not have a verified context window recorded for this model — confirm the limit in OpenAI's documentation before designing long-repository workflows
- Throughput and time-to-first-token benchmarks are not yet populated in our database, so latency behavior cannot be compared numerically here
- No published benchmark scores (for example SWE-bench style coding evaluations) are recorded in our metadata for this entry
- Modality support beyond text is not confirmed in our data, so do not assume image or audio input
- Codex-tuned models are optimized around coding tasks, which may make them a less obvious default for open-ended conversational or creative workloads
Key Features
About GPT-5.2 Codex
Common Use Cases
GPT-5.2 Codex fits software engineering workloads: generating new code from a specification, refactoring existing modules, explaining unfamiliar code, drafting tests, and diagnosing failing builds. Its Codex positioning makes it a candidate for coding agents that iterate — read files, propose a patch, check the result, revise — and for developer tooling such as IDE completions, pull request review bots, and migration scripts that need syntactically valid output across languages. Because we do not have a confirmed context window on file, teams planning whole-repository ingestion should verify the limit against OpenAI's documentation and, where the codebase is large, pair the model with retrieval or file-scoped chunking rather than assuming a single oversized prompt will fit.
Frequently Asked Questions
How much does GPT-5.2 Codex cost to use?
Pricing depends on the provider and on the pricing type — input versus output tokens, and any batch or cached-input rates a given provider offers. Rates also change over time. Check the pricing table on this page for the current per-token figures from each provider that carries this model.
What is GPT-5.2 Codex best used for?
It is positioned for coding work: writing and refactoring code, reviewing diffs, generating tests, and powering agentic development loops in IDEs or CI tooling. Codex-branded models in OpenAI's lineup are tuned toward software engineering rather than general-purpose conversation.
What context window does GPT-5.2 Codex support?
We do not have a confirmed context window recorded for this model in our database. Consult OpenAI's model documentation for the authoritative token limit before building workflows that depend on large inputs.
Does GPT-5.2 Codex accept images or other non-text input?
Our metadata does not confirm any modality beyond text for this entry, so we can't state what non-text inputs are supported. Verify supported input types with OpenAI's documentation for the specific model version you plan to call.
How does GPT-5.2 Codex differ from other GPT-5.x models?
The Codex suffix marks it as the coding-oriented variant in the series, aimed at software engineering and code-editing tasks, whereas the non-Codex GPT-5.x models are positioned as general-purpose chat models. Because both are served through the same chat interface, the practical way to choose is to run your own coding prompts against each and compare quality alongside the rates in the pricing table above.